Design of protein-protein interfaces

Message boards : Rosetta@home Science : Design of protein-protein interfaces

To post messages, you must log in.

1 · 2 · 3 · 4 . . . 6 · Next

AuthorMessage
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 56754 - Posted: 7 Nov 2008, 7:08:03 UTC

Hello,

My name is Sarel Fleishman and I've been a postdoc in the Baker lab for the past two years. My project deals with structure prediction and design of protein-protein interfaces and you may have read a few messages from me on CAPRI and structure prediction. Now, I'm very excited to tell you that with minirosetta v1.40 we can do design work using the massive computational power of Rosetta @ Home.

A few words on protein-interface design. The primary challenge in this field is to be able to take a protein target and to design another protein that would bind to it in a specific way. Nature provides us with hundreds of thousands of examples of such protein-binding events. Such events are used for the amplification of signals within and between our cells in processes related to growth and development as well as for recognition, e.g., in the immune system. When such signals go awry, protein interactions become the center of events for uncontrolled cell growth, or cancer. Many pathogenic bacteria and viruses hijack molecular recognition processes to promote their growth and proliferation causing sickness and endangering lives.

These processes being so central to both health and disease it is hardly surprising that being able to manipulate them computationally is a major ongoing goal of molecular biology. Being able to target a protein and bind to it would open the way to novel therapies for a large number of diseases.

We have selected a small number of protein targets for which we want to design protein binders. As an example, one such target that I have been working on is cholera toxin. This protein is a crucial component of the process by which the cholera bacterium causes cells in the gut to excrete large amounts of water, which causes death from dehydration (see the following wiki page for more details: http://en.wikipedia.org/wiki/Cholera). We have developed a computational strategy that allows us to design proteins to bind to the cholera toxin and disable it.

We are now in the process of testing this and similar design methodologies on a number of other targets. But expanding the number of targets we quickly realized that we need a lot more computational resources to adequately address this problem. This is why we have turned to Rosetta @ Home and to you for your help in this exciting project.

The simulations that you will see in protein interface design will be quite different from one another. In each case we tailor the design strategy to the particular protein target, stressing, for instance, the formation of interactions with a specific key region on the protein surface. In general, though, the simulations will involve docking steps, where the protein binder moves with respect to the target and design, where amino acid residues on the surface of the protein binder change in order to better attach to the target. Promising protein designs are synthesized in our lab and tested for binding to the target protein.

I am working on this project with my colleagues Eva and Jacob, and for each target that we test using Rosetta @ Home we will provide background material on the target and why we selected it on this thread.

These design simulations tend to use more memory than many prediction runs (typically at most 800Mb). We will test different ways of reducing this memory load so that our simulations could run on all participating computers in the Rosetta @ Home project, but initially we will only run these simulations on computers that can handle this memory restriction. Please report any problems that you might have with this simulation.

Thank you very much for participating in this project! I'm looking forward to getting feedback and results from you.
ID: 56754 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jcorn

Send message
Joined: 27 Jan 06
Posts: 6
Credit: 198,437
RAC: 0
Message 56763 - Posted: 7 Nov 2008, 18:05:54 UTC

My name is Jacob Corn, and I'm a postdoc in the Baker Lab, working with Eva and Sarel on the design of protein-protein interfaces. As Sarel mentioned, this is an incredibly important problem, both for basic science and molecular medicine. But it is also an incredibly difficult problem. Without your support, we would never have enough computing power to work on these kinds of projects.

One target that I am attempting to design towards is interleukin 23 (aka IL23). Your body normally uses this protein to trigger the inflammatory response, which is a normal part of your body's immune system. However, if IL23 messages start to run out of control, they can cause serious autoimmune disorders, such as Chron's Disease. You can find more information on IL23 here
http://en.wikipedia.org/wiki/Interleukin_23
and Chron's Disease here
http://en.wikipedia.org/wiki/Chron%27s_disease.

This week you may receive jobs with titles like "IL23p40_p40BrubYhbond_design_jecorn". These design simulations first dock two proteins against each other (one of which is IL23), then try to iteratively optimize the surface of one protein to better match the surface of IL23. Using the results of your calculations, I will the designed proteins and test them for binding to IL23, hopefully producing a new inhibitor to combat runaway diseases caused by runaway IL23 signals.

I'm very excited about this project, and look forward to your feedback.
ID: 56763 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile proxima

Send message
Joined: 9 Dec 05
Posts: 44
Credit: 4,148,186
RAC: 0
Message 56778 - Posted: 8 Nov 2008, 22:17:33 UTC - in response to Message 56763.  

I'm very excited about this project, and look forward to your feedback.


It's been great getting some new 'science' news from your and the rest of the project team. This sort of stuff is *really* motivating to a lot of us - please keep the news coming, and good luck with your project.
Alver Valley Software Ltd - Contributing ALL our spare computing power to BOINC, 24x365.
ID: 56778 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Michael G.R.

Send message
Joined: 11 Nov 05
Posts: 264
Credit: 11,247,510
RAC: 0
Message 56795 - Posted: 9 Nov 2008, 22:14:08 UTC - in response to Message 56778.  

It's been great getting some new 'science' news from your and the rest of the project team. This sort of stuff is *really* motivating to a lot of us - please keep the news coming, and good luck with your project.


I second that.
ID: 56795 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Naesbye

Send message
Joined: 30 Jul 08
Posts: 5
Credit: 201,436
RAC: 0
Message 56803 - Posted: 10 Nov 2008, 13:24:41 UTC

Me too. Having a fundamental idea of what we're helping out with, and knowing that it serves a beneficial purpose is very inspiring.
ID: 56803 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 113,931,835
RAC: 64,055
Message 56805 - Posted: 10 Nov 2008, 16:28:47 UTC

fourthed - this info gets posted around various team forums - i'll post it on the XPC forum at some point as a lot of them don't read these forums but do read their team forums. Keeps everyone interested and often gets people to ramp up productivity.
ID: 56805 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 56855 - Posted: 11 Nov 2008, 21:31:12 UTC - in response to Message 56754.  

Hello,

My name is Sarel Fleishman and I've been a postdoc in the Baker lab for the past two years. My project deals with structure prediction and design of protein-protein interfaces and you may have read a few messages from me on CAPRI and structure prediction. Now, I'm very excited to tell you that with minirosetta v1.40 we can do design work using the massive computational power of Rosetta @ Home.

These design simulations tend to use more memory than many prediction runs (typically at most 800Mb). We will test different ways of reducing this memory load so that our simulations could run on all participating computers in the Rosetta @ Home project, but initially we will only run these simulations on computers that can handle this memory restriction. Please report any problems that you might have with this simulation.

Thank you very much for participating in this project! I'm looking forward to getting feedback and results from you.


Sounds like a good idea, but we need better estimates of how long these workunits will run and how much memory they will take. I've had as much memory as my motherboard can handle (2 GB for a CPU with 2 cores) for some time, but am now getting into the same slowdown of regular tasks that persuaded me to buy that much memory. Also, some of the minirosetta w1.40 workunits, typically those with 4704 in their names, are running much longer than my preferred 6 hours, without giving any workunits from other BOINC projects a chance to run, and then giving a poor credits to CPU time required ratio.

Could they be set up to make more frequent checkpoints?

Also, do you know where to find a program that can measure both the parts of workunits in progress in physical memory and in virtual memory, for Vista SP1?
ID: 56855 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 113,931,835
RAC: 64,055
Message 56856 - Posted: 11 Nov 2008, 21:50:30 UTC - in response to Message 56855.  

Could they be set up to make more frequent checkpoints?
unfortunately not - it's been discussed a number of times with the same conclusion - there are only certain points that are suitable for checkpointing.


Also, do you know where to find a program that can measure both the parts of workunits in progress in physical memory and in virtual memory, for Vista SP1?
I think process explorer works on Vista(?)

ID: 56856 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Chilean
Avatar

Send message
Joined: 16 Oct 05
Posts: 711
Credit: 26,694,507
RAC: 0
Message 56888 - Posted: 13 Nov 2008, 0:05:13 UTC

R@H Web designer should post this on the homepage and make it look neat and professional so people who come to the website for the first time can say "wow, this is for real" and join and not LEAVE.

Just an opinion.
ID: 56888 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
jcorn

Send message
Joined: 27 Jan 06
Posts: 6
Credit: 198,437
RAC: 0
Message 56916 - Posted: 13 Nov 2008, 20:01:31 UTC

Hi everyone,
A quick update on computational design in general, and IL23 in particular.

As many people noticed, these design jobs are taking longer and using more memory than most R@H work units. This is because we are working with proteins that are much larger than what you normally fold during an abinitio run. For reference, imagine that abinitio runs are folding a steel chain with 1-inch links, and that the position of each link changes during the course of the workunit. Abinitio workunits then origami-fold about 8 feet of constant chain into a nice, pretty structure.

By contrast, our design workunits are dealing with two chains, each of which is twenty five to fifty feet long. Not only can the whole folded chain move, the positions of each link can shift slightly, *and* we're constantly swapping out links as the simulation progresses! We have twenty different links to play with, and each position on each chain could be any one of the twenty. Our challenge is to figure out how the two chains should lay on top of one another and what link should be at what position to make the the chains best fit together. As you can imagine, all of that takes a great deal of computing power, and about half a gigabyte of memory. We've put a temporary hold on future design workunits until we do a more benchmarking and fix the issues that you've brought up. For example, we are considering restricting design runs to machines with no less than 1GB of RAM.

Now the good news: Thanks to all of your work, the IL23 design project is moving at an incredible pace. Over the last 5 days, you generated half a million potential therapeutics for IL23! That's completely unprecedented for this kind of project, and would have taken me forever to do on our lab computers. So even though it was a somewhat rocky start, you have all made a very significant contribution towards creating an IL23 anti-inflammatory drug. I'm very much looking forward to sifting through all of the potential therapeutics you've designed, and I feel that R@H really holds the key to making this project work.
ID: 56916 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 56986 - Posted: 16 Nov 2008, 8:05:26 UTC - in response to Message 56856.  
Last modified: 16 Nov 2008, 8:06:59 UTC



Also, do you know where to find a program that can measure both the parts of workunits in progress in physical memory and in virtual memory, for Vista SP1?
I think process explorer works on Vista(?)


dcdc,

How do I even find process explorer? When I did a help file search for process explorer, it gave me at least 60 help file pages to check for more information, but none of the first 30 say anything about whether Vista even includes a program by that name, much less how to start it.
ID: 56986 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 56987 - Posted: 16 Nov 2008, 8:28:06 UTC - in response to Message 56916.  

Hi everyone,
A quick update on computational design in general, and IL23 in particular.

By contrast, our design workunits are dealing with two chains, each of which is twenty five to fifty feet long. Not only can the whole folded chain move, the positions of each link can shift slightly, *and* we're constantly swapping out links as the simulation progresses! We have twenty different links to play with, and each position on each chain could be any one of the twenty. Our challenge is to figure out how the two chains should lay on top of one another and what link should be at what position to make the the chains best fit together. As you can imagine, all of that takes a great deal of computing power, and about half a gigabyte of memory. We've put a temporary hold on future design workunits until we do a more benchmarking and fix the issues that you've brought up. For example, we are considering restricting design runs to machines with no less than 1GB of RAM.



Do machines with more than one CPU core, but no more memory than 1 GB times the number of CPU cores, qualify? Mine, for instance. If so, is there something to prevent all the CPU cores from trying to run such workunits at the same time? I don't mind the extra time required, so long as there are enough checkpoints to restart after Vista updates and similar causes for a reboot
without losing too much CPU time. I would, however, expect accurate enough estimates of the time and memory required to give me a reasonable amount of credits and to help decide when my machine can start such a workunit without slowing it down too much.

ID: 56987 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile dcdc

Send message
Joined: 3 Nov 05
Posts: 1829
Credit: 113,931,835
RAC: 64,055
Message 56989 - Posted: 16 Nov 2008, 11:11:46 UTC - in response to Message 56986.  
Last modified: 16 Nov 2008, 11:14:38 UTC



Also, do you know where to find a program that can measure both the parts of workunits in progress in physical memory and in virtual memory, for Vista SP1?
I think process explorer works on Vista(?)


dcdc,

How do I even find process explorer? When I did a help file search for process explorer, it gave me at least 60 help file pages to check for more information, but none of the first 30 say anything about whether Vista even includes a program by that name, much less how to start it.

http://technet.microsoft.com/en-us/sysinternals/bb896653.aspx
ID: 56989 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 56993 - Posted: 16 Nov 2008, 13:46:43 UTC

I found a number of web sites refering to IL23.

IL-23: a master regulator in Crohn disease
http://www.nature.com/nm/journal/v13/n1/full/nm0107-26.html

Major Genetic Link to Crohn's and Colitis Found
http://www.ccfa.org/about/press/il23

Production of IL12p70 and IL23 by monocyte-derived dendritic cells in children with...
http://gut.bmj.com/cgi/content/full/57/10/1480?rss=1

IL-23 drives a pathogenic T cell population that induces autoimmune inflammation
http://jem.rupress.org/cgi/content/abstract/201/2/233

IL23 (Cytokines & Cells Online Pathfinder Encyclopaedia)
http://www.copewithcytokines.de/cope.cgi?key=IL23

Interleukin-23 promotes a distinct CD4 T cell activation state characterized by the production of interleukin-17
http://www.ncbi.nlm.nih.gov/pubmed/12417590

Interleukin 23 - Wikipedia
http://en.wikipedia.org/wiki/Interleukin_23

Looks like it is imvolved in at least some of the autoimmune diseases.
ID: 56993 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 57144 - Posted: 21 Nov 2008, 19:58:55 UTC
Last modified: 21 Nov 2008, 21:10:49 UTC

There has been a lot of interest in this thread, so I thought that I would write a little more about our design strategy and the importance of Rosetta @ Home's computational resources for the project's success. First, our approach in designing a new protein inhibitor to a natural target revolves around the idea that nature provides us with many thousands of diverse examples of protein interactions from which to learn. Studying these interactions in detail, we find that there are many unifying principles, such as the importance of energetic considerations (for more information, see http://en.wikipedia.org/wiki/Van_der_Waals_force). But, beyond these unifying principles it is striking to see just how diverse a phenomenon protein-protein interactions is! A great resource to learn more about the beauty and the implications of protein structures and interactions is the RCSB's molecule of the month feature, and here's one such feature on one of my targets, cholera toxin:
http://www.rcsb.org/pdb/static.do?p=education_discussion/molecule_of_the_month/pdb69_1.html
As these illustrations and many other examples show, many protein interactions are uniquely optimized to take advantage of the physical interactions between chemical groups on the proteins' surfaces. Additionally, the surfaces of the proteins usually show very high shape complementarity as in the spectacular example provided by the protein that constitutes the majority of our connective tissue, collagen:
http://www.rcsb.org/pdb/static.do?p=education_discussion/molecule_of_the_month/pdb4_1.html

Building on the idea that Rosetta captures correctly the unifying energetic principles, as evidenced by its successes in structure prediction and many different protein design tasks, we are still faced with the daunting challenge of how to best capture the great diversity encompassed by the protein-interaction phenomenon. In other words, how can we, for instance, find a protein, the surface of which would ideally fit with the surface of our target of choice as in the collagen example above? This is exactly the point where we need all that sampling power that you make available through the Rosetta @ Home project. Given a target protein, such as cholera toxin, we would like to massively test different amino-acid sequences. We would then like to make sure that each such sequence makes good physical sense, shows high shape complementarity, and would provide sufficient affinity. Since each target implies its own set of unique challenges (e.g., its molecular shape is unique), we specifically tailor a different protocol (or sequence of design steps) to each target and we want to test several such protocols for each given target in a trial-and-error way. Multiply all these computationally intensive tasks by the many millions of conformations that we would like to test for our potential binders with respect to the target and you see how it is that only through Rosetta @ Home we can begin to imagine finding a comprehensive solution to this great challenge!
ID: 57144 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 57154 - Posted: 22 Nov 2008, 0:55:53 UTC

ID: 57154 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 57155 - Posted: 22 Nov 2008, 1:20:41 UTC

ID: 57155 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile robertmiles

Send message
Joined: 16 Jun 08
Posts: 1223
Credit: 13,806,125
RAC: 3,336
Message 59423 - Posted: 7 Feb 2009, 17:16:37 UTC - in response to Message 56856.  

Also, do you know where to find a program that can measure both the parts of workunits in progress in physical memory and in virtual memory, for Vista SP1?
I think process explorer works on Vista(?)
[/quote]

It does, but it took a few months to find where it went when I installed it.

It appears to be showing me that BOINC is still using both cores of my CPU at 100%, even though I first tried to lower this to 98% and then 90% in order to help test for certain problems over on RALPH@home.
ID: 59423 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Profile Sarel

Send message
Joined: 11 May 06
Posts: 51
Credit: 81,712
RAC: 0
Message 64838 - Posted: 7 Jan 2010, 0:35:29 UTC

Hello,

Since I last wrote (a long time ago) we've made substantial progress in the design of protein interactions. We have developed a new methodology and tested it extensively in recapitulating known protein interactions as well as in the design of new interactions. In all of our tests the results have been very favourable, and most importantly, we now have a handful of designs that show specific binding towards their targets in experiments!

We're obviously very excited about the potential applications of this methodology for the design of inhibitors and binders of drug targets and other biomedically interesting proteins. The target that has most excited us is hemagglutinin from pandemic influenza strains, including of the swine and Spanish flu. This protein is a prime drug candidate and its structure with an antibody that neutralizes these strains of influenza has been solved. We are using our new methodology in order to compute binders of hemagglutinin that may be easier to mass produce than is the antibody. As might be expected, the complexity of designing a binder towards hemagglutinin surpasses that of other protein targets that we had previously attempted and therefore requires substantially more computational resources. We are hoping that with your contributions to ROSETTA@HOME we may obtain many putative inhibitors for experimental testing.

We have other targets in our pipeline that we are very excited about. The design simulations that we will send out will contain a description of the target protein, so for instance, the simulations pertaining to hemagglutinin will be labeled "design of an inhibitor of influenza hemagglutinin". We'll also update this thread with descriptions of other targets and the results of the simulations and experimental characterization. We're excited to see how many new solutions to this problem you will come up with!
ID: 64838 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
Michael G.R.

Send message
Joined: 11 Nov 05
Posts: 264
Credit: 11,247,510
RAC: 0
Message 64839 - Posted: 7 Jan 2010, 1:28:42 UTC

Thank you for the update, Sarel. Sounds like promising developments.
ID: 64839 · Rating: 0 · rate: Rate + / Rate - Report as offensive    Reply Quote
1 · 2 · 3 · 4 . . . 6 · Next

Message boards : Rosetta@home Science : Design of protein-protein interfaces



©2024 University of Washington
https://www.bakerlab.org